• directed acyclic graphs
• SVAR
• VAR
• time-series
• causality
• causal discovery

This study presents a novel algorithm for causal discovery in time series data. Our approach distinguishes itself as a highly efficient and effective alternative to the current state-of-the-art methods within the same framework. Efficiency is crucial in causal discovery, particularly in applications involving high-dimensional datasets, where repeated training procedures are often necessary for tuning or verifying the stability of the inferred causal structure. Our algorithm is rooted in the Structural Vector Autoregressive (SVAR) model framework, effectively tackling two primary challenges in causal discovery from time series data: the sampling process being slower than the time scale and the inference of causality from a single time series.